Upstage Solar Delegation

Delegate long text generation to Upstage Solar Pro3 while the primary model focuses on planning, reasoning, and tool calls.

Call Routes

Upstage Direct (Recommended)

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["UPSTAGE_API_KEY"],
    base_url="https://api.upstage.ai/v1"
)

response = client.chat.completions.create(
    model="solar-pro3",
    messages=[{"role": "user", "content": "Write a detailed report..."}]
)
print(response.choices[0].message.content)

Get your key at console.upstage.ai.

Via OpenRouter

Model reference: https://openrouter.ai/upstage/solar-pro-3

Enabled Sessions

Check runtime/session settings to see where delegation is enabled.

Typical values:

• main: direct/main session
• slack: Slack messages
• telegram: Telegram messages

If the current session is not enabled, skip delegation and respond directly.

Delegation Threshold

Use a minimum output-token threshold:

• estimated output tokens >= threshold → delegate to Solar
• estimated output tokens < threshold → respond directly

Common examples:

• 0: delegate all responses
• 200: delegate medium/long responses only

If no threshold is configured, use 200 as default.

How to Delegate

1. Estimate output token length.
2. Check whether current session is enabled.
3. If enabled and estimate >= threshold, run sessions_spawn with Solar Pro3.

Example spawn task pattern:

<context + writing instruction>

Important: Do not use any tools. Output text only. No file read/write or tool calls allowed.

Set model to:

• solar-pro3 (Upstage direct — recommended)
• openrouter/upstage/solar-pro-3 (via OpenRouter)

Main Session

Use sessions_spawn and wait for auto-announced result.

Messenger Sessions (Slack/Telegram/etc.)

Use sessions_spawn, then fetch the final assistant text and forward it through the appropriate message channel/thread.

If no final text arrives within a reasonable timeout, fall back to direct response.

Output Files

• Inline default: when the user expects the response in chat, do NOT write a file — return the text directly.
• When saving to disk: write to <system-temp>/<timestamp>-solar.<ext> (e.g., /tmp/20260504-153022-solar.md). Pick .md for markdown content, .txt for plain prose.
• Override: if the user specifies an output path, use it.
• Always print the resolved absolute path when writing to disk.

Rules

• Pass through Solar output as-is (no extra summary/footer/metadata).
• Keep non-user-facing orchestration text minimal.
• Do not output intermediate narration between tool calls.
• Always include the “no tool call” instruction in spawn tasks.

Decision Guide

Delegate when above threshold and user expects substantial writing:

• explanations
• summaries
• reports
• long-form answers

Keep direct response for:

• short operational confirmations
• urgent low-latency replies
• responses that must include immediate tool-call outputs

Configuration Changes

Users may request:

• threshold changes (e.g., “set threshold to 300”)
• session enable/disable (e.g., enable delegation in Slack)

Apply updates to persistent memory/config used by your environment.

First-Time Setup

Option A: Upstage Direct (Recommended)

1. Confirm user wants setup.
2. Verify UPSTAGE_API_KEY environment variable is configured. Get key at console.upstage.ai.
3. Set base_url to https://api.upstage.ai/v1, model to solar-pro3.
4. Confirm delegation is active and report current threshold.

Option B: Via OpenRouter

1. Confirm user wants setup.
2. Confirm OpenRouter API key is available (OPENROUTER_API_KEY).
3. Add OpenRouter provider + Solar model via gateway config update.
4. Restart/reload gateway as required.
5. Confirm delegation is active and report current threshold.

For manual setup details, see references/setup-guide.md.